Chapter 3 Geometry of convex functions - Meboo Publishing ...
Chapter 3 Geometry of convex functions - Meboo Publishing ...
Chapter 3 Geometry of convex functions - Meboo Publishing ...
Create successful ePaper yourself
Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.
3.8. MATRIX-VALUED CONVEX FUNCTION 257<br />
3.8.3 second-order condition, matrix function<br />
The following line theorem is a potent tool for establishing <strong>convex</strong>ity <strong>of</strong> a<br />
multidimensional function. To understand it, what is meant by line must first<br />
be solidified. Given a function g(X) : R p×k →S M and particular X, Y ∈ R p×k<br />
not necessarily in that function’s domain, then we say a line {X+ t Y | t ∈ R}<br />
passes through domg when X+ t Y ∈ domg over some interval <strong>of</strong> t ∈ R .<br />
3.8.3.0.1 Theorem. Line theorem. [59,3.1.1]<br />
Matrix-valued function g(X) : R p×k →S M is <strong>convex</strong> in X if and only if it<br />
remains <strong>convex</strong> on the intersection <strong>of</strong> any line with its domain. ⋄<br />
Now we assume a twice differentiable function.<br />
3.8.3.0.2 Definition. Differentiable <strong>convex</strong> matrix-valued function.<br />
Matrix-valued function g(X) : R p×k →S M is <strong>convex</strong> in X iff domg is an<br />
open <strong>convex</strong> set, and its second derivative g ′′ (X+ t Y ) : R→S M is positive<br />
semidefinite on each point <strong>of</strong> intersection along every line {X+ t Y | t ∈ R}<br />
that intersects domg ; id est, iff for each and every X, Y ∈ R p×k such that<br />
X+ t Y ∈ domg over some open interval <strong>of</strong> t ∈ R<br />
Similarly, if<br />
d 2<br />
dt 2 g(X+ t Y ) ≽ S M +<br />
0 (611)<br />
d 2<br />
dt g(X+ t Y ) ≻ 0 (612)<br />
2<br />
S M +<br />
then g is strictly <strong>convex</strong>; the converse is generally false. [59,3.1.4] 3.21 △<br />
3.8.3.0.3 Example. Matrix inverse. (confer3.4.1)<br />
The matrix-valued function X µ is <strong>convex</strong> on int S M + for −1≤µ≤0<br />
or 1≤µ≤2 and concave for 0≤µ≤1. [59,3.6.2] In particular, the<br />
function g(X) = X −1 is <strong>convex</strong> on int S M + . For each and every Y ∈ S M<br />
(D.2.1,A.3.1.0.5)<br />
d 2<br />
dt 2 g(X+ t Y ) = 2(X+ t Y )−1 Y (X+ t Y ) −1 Y (X+ t Y ) −1 ≽<br />
3.21 The strict-case converse is reliably true for quadratic forms.<br />
S M +<br />
0 (613)